Figure 5

Integrative prediction of known genetic interactions. (A) positive interactions, (B) negative interactions in the SGA - GIM data pair. The true positive rate (TPR or sensitivity) is the fraction of mutant pairs correctly classified into its true interaction class, and the false positive rate (FPR, or 1 - specificity) is the fraction of non-interacting gene pairs incorrectly classified into the particular interaction class. The dotted trace illustrates the average performance of a random classifier. The colour traces indicate the different rank aggregation functions (minimum, maximum, product, and Borda count), which combine the interaction scores across datasets, compared to the ranking of the interaction scores using a single dataset alone (SGA rank, GIM rank, and E-MAP rank). Interaction scores were based on the fixed QMA setting for scoring positive interactions. The minimum function was omitted from the positive interactions and the maximum from the negative interactions for the clarity of illustration. The performance metrics of all the functions are summarized in Table 4.